Improving ASTRO's Perception Capability for a more Natural Follow-Me Task

Bruno Ferreira, Alessandra Sorrentino, Paulo Menezes, Jorge Batista, Jorge Dias, Laura Fiorini, Filipo Cavallo

    Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

    1 Scopus citations

    Abstract

    Social assistive robots usually encompass a great compromise between the advanced perception models that one can use and their computing capabilities. The ideal approaches are always oriented towards low power consumption while maintaining a higher order of responsiveness to the surroundings. Therefore, we present in this paper, an improvement of the follow-me system on ASTRO. In detail, we propose the use of Mediapipe SDK for human detection and tracking, when ASTRO is meant to accompany someone that is walking. A study on the new velocity and distance profiles that the robot keeps from the individuals is presented and we have also evaluated how it affects their perception of being safe. The presented results show that our new approach allows the system to achieve real-time performance by becoming ≈ 9.1× faster, smoothing, and keeping a more natural distance from the user.

    Original languageBritish English
    Title of host publication6th Experiment at International Conference, exp.at 2023 - Proceedings
    EditorsAlberto Cardoso, Maria Teresa Restivo
    PublisherInstitute of Electrical and Electronics Engineers Inc.
    Pages79-84
    Number of pages6
    ISBN (Electronic)9798350314403
    DOIs
    StatePublished - 2023
    Event6th Experiment at International Conference, exp.at 2023 - Evora, Portugal
    Duration: 5 Jun 20237 Jun 2023

    Publication series

    Name6th Experiment at International Conference, exp.at 2023 - Proceedings

    Conference

    Conference6th Experiment at International Conference, exp.at 2023
    Country/TerritoryPortugal
    CityEvora
    Period5/06/237/06/23

    Keywords

    • assistive systems
    • human-robot interaction
    • social robotics

    Fingerprint

    Dive into the research topics of 'Improving ASTRO's Perception Capability for a more Natural Follow-Me Task'. Together they form a unique fingerprint.

    Cite this